bioRxiv preprint doi: https://doi.org/10.1101/2020.12.18.423523; this version posted December 21, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

1 Journal of Experimental Biology - Research Article 2 Cactus mouse circadian metabolism 3 4 Disentangling environmental drivers of circadian metabolism in desert-adapted mice 5 6 Authors

7 Jocelyn P. Colella1✝, Danielle M. Blumstein1, Matthew D. MacManes1*

8 9 Affiliations 10 1University of New Hampshire, Molecular, Cellular, and Biomedical Sciences Department, 11 Durham, NH 03824 (JPC [email protected]; DMB [email protected]; MDM 12 [email protected]) 13 ✝University of Kansas, Biodiversity Institute, Lawrence, KS 66045

14 15 *corresponding author: [email protected] 16 17 KEYWORDS: circadian torpor, deer mice, lipogenesis, metabolism, Peromyscus eremicus, 18 physiology 19 20 SUMMARY STATEMENT 21 Continuous metabolic phenotyping of desert-adapted cactus mice (Peromyscus eremicus) 22 identifies significant metabolic differences between the sexes and circadian patterning 23 consistent with lipogenesis and environmental entrainment. 24 25 ABSTRACT 26 Metabolism is a complex phenotype shaped by natural environmental rhythms, as well as 27 behavioral, morphological, and physiological adaptations. Although historically studied under 28 constant environmental conditions, continuous metabolic phenotyping through environmental 29 transitions now offers a window into the physiological responses of organisms to changing 30 environments. Here, we use flow-through respirometry to compare metabolic responses of the 31 desert-adapted cactus mouse (Peromyscus eremicus) between diurnally variable and constant 32 environmental conditions. We contrast metabolic responses to circadian cycles in photoperiod, 33 temperature, and humidity, against those recorded under constant hot-and-dry and constant

1 bioRxiv preprint doi: https://doi.org/10.1101/2020.12.18.423523; this version posted December 21, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

34 cold-and-wet conditions. We found significant sexual dimorphism in metabolic responses, 35 despite no measurable difference in body weight. Males seem to be more heat tolerant and

36 females more cold tolerant. Under circadian environmental cycling, the ratio of CO2 produced to

37 O2 consumed (the respiratory quotient or respiratory exchange ratio) reached greater than one, 38 a pattern that strongly suggests that lipogenesis is contributing to the production of energy and 39 endogenous water in this species. This hypothesis is consistent with the results of previous 40 dehydration experiments in this species, which documented significant weight loss in response 41 to dehydration, without other physiological impairment. Our results are also consistent with 42 historical descriptions of circadian torpor in this species (torpid by day, active by night), but 43 reject the hypothesis that this pattern is driven by food restriction or negative water balance, as 44 both resources were available to animals throughout the experiments. 45 46 INTRODUCTION 47 Our planet is undergoing unprecedented environmental change, which is exerting new, strong 48 selective pressures on species, forcing them to relocate, adapt in situ, or risk extinction. 49 Warming temperatures and increased desertification are predicted to impact most species in 50 North America (Parmesan and Yohe, 2003; Urban, 2015; IPCC, 2018). One productive line of 51 research has been related to desert animals (Greenville et al., 2017; Kordonowy et al., 2017; 52 MacManes, 2017; Riddell et al., 2019) which are already adapted to hot and dry conditions and 53 therefore may illustrate mechanisms (e.g., molecular, behavioral, or physiological) through 54 which animals may adapt to a warming climate (Vale and Brito, 2015; Colella et al., in press). 55 Much research has been devoted to understanding genetic correlates with heat- and 56 dehydration-tolerance in extant desert-adapted taxa. Of equal importance, are the physiological 57 adaptations of desert species, which are predicted to play an important role in species 58 responses to climate change (Kearney et al., 2009; Boyles et al., 2011; Huey et al., 2012). Yet, 59 adaptive thermoregulatory mechanisms do not evolve under fixed (or static) environmental 60 conditions, as they have traditionally been measured (Hulbert and Else, 2004). Instead, 61 thermoregulatory responses are also shaped by natural environmental rhythms, including 62 circadian and seasonal variation in temperature, humidity, and photoperiod, among other 63 variables (Morrison et al., 2008; Seebacher, 2009). Circadian rhythms are observed in most 64 mammals and are in part governed by genetics, as well as the interplay between the perception 65 of environmental change and neuroendocrine-mediated physiological responses (Miranda- 66 Anaya et al. 2019). Examination of the physiological responses of organisms to both static and 67 dynamic environmental conditions provides a window into the flexibility of thermoregulatory

2 bioRxiv preprint doi: https://doi.org/10.1101/2020.12.18.423523; this version posted December 21, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

68 responses to changing environments and will be increasingly important to understanding, and 69 ultimately, predicting (Garland and Carter, 1994; Dalziel et al., 2009; Easterling et al., 2000; 70 IPCC, 2012; Huey et al., 2012) species responses to climate change. 71 Deer mice of the genus Peromyscus inhabit an exceptional range of thermal 72 environments in North America, ranging from cold high elevation mountain tops to arid hot 73 deserts (Hock, 1964; Cheviron et al., 2014; MacManes and Eisen, 2014; Harris and Munshi- 74 South, 2017; Storz et al., 2019). Under projected climate scenarios (e.g., IPCC, 2018), 75 adaptations that increase fitness in warmer, drier environments are predicted to be increasingly 76 favored (Merkt and Taylor, 1994; Gienapp et al., 2008; Kingsolver, 2009). Cactus mice 77 (Peromyscus eremicus) are endemic to the southwestern deserts of North America and exhibit 78 a number of morphological (large ears, large surface to volume ratio), behavioral (saliva 79 spreading, Murie, 1961), and physiological (low metabolism, Murie, 1961, McNab and Morrison, 80 1963; anuria, Kordonowy et al., 2017) adaptations to high heat, low water desert environments. 81 The suitability of this species to life in captivity also makes them a promising experimental 82 model for identifying and characterizing thermoregulatory mechanisms. 83 In addition to being able to withstand environmental extremes, desert taxa, including 84 cactus mice, are also adapted to large circadian fluctuations in environmental conditions. In the 85 Sonoran Desert, daily temperatures can swing between -4ºC to 38ºC (Reid, 1987; Sheppard et 86 al., 2002), with weeks or even months passing between rainfall events. Water is required for 87 basic physiological functioning (Haussinger, 1996; Montain et al., 1999), but in desert 88 environments water loss often exceeds intake (Heimeier et al., 2002) and evaporative water 89 loss is further exacerbated by extreme heat and aridity (Cheuvront et al., 2010). Most 90 endotherms are limited in their ability to metabolically respond to sub-optimal environmental 91 conditions (Bennet and Ruben, 1979; Fuglei and Oritsland, 1999), with a notable exception 92 being those that can enter hibernation or temporary torpor (Ruf and Geiser, 2015). In response 93 to water stress, cactus mice enter seasonal torpor, reducing above ground activity and 94 depressing metabolism during the dry season (Macmillen, 1965). Early investigations found that 95 torpor in cactus mice could be initiated at any time by food restriction, but could only be 96 triggered by dehydration during the summer months, which suggests some degree of metabolic 97 entrainment (Macmillen, 1965, 1972). As a nocturnal species, cactus mice also reduce activity 98 levels during the daytime, seeking cooler, shaded parts of their habitat, hiding underground, or 99 even entering circadian torpor during the day (torpid by day, active by night) to decrease heat 100 load, respond to resource limitation, maximize heat dissipation, and minimize respiratory water 101 loss (Macmillen, 1965, 1972; Degen, 2012; Alonso et al., 2016). Ultimately, a balance exists

3 bioRxiv preprint doi: https://doi.org/10.1101/2020.12.18.423523; this version posted December 21, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

102 between environmental heterogeneity (e.g., food availability, thermal stress), which requires 103 animals to sense and respond to external conditions in real time via reactionary mechanisms, 104 and environmental predictability which can lead to the evolution of anticipatory biorhythms 105 based on regular abiotic conditions (e.g., photoperiod). 106 Here, we examine patterns of metabolic variation in cactus mice across a 24-hour period 107 in response to both diurnally variable and constant environments. We hypothesize that cactus 108 mice will exhibit physiological responses consistent with desert adaptation, that minimize 109 daytime energy expenditure and water loss. We further hypothesize that cactus mouse 110 metabolism will be well-tuned to circadian patterns of environmental variation, and that circadian 111 patterning will be disrupted under constant environmental treatments. Comparing metabolic 112 responses between sexes and across constant and environmentally variable treatment groups, 113 we show that (i) males are more heat tolerant than females, (ii) patterns of metabolic variation 114 are well-tuned to environmental cycling, and (iii) lipogenesis offers a potential mechanism of 115 endogenous water production, only employed under variable environmental conditions. 116 117 MATERIALS AND METHODS 118 Mice & environmental conditions 119 Mice were cared for, handled, and sampled in accordance with the University of New 120 Hampshire’s Institutional Animal Care and Use Committee (AUP #180705). All mice used in this 121 study were bred from wild-derived lines maintained at the Peromyscus Genetic Stock Center of 122 the University of South Carolina (Columbia, South Carolina, USA). All mice were sexually 123 mature, non-reproductive adults between three and nine months of age. Animals were fed 124 LabDiet 5015* (19% protein, 26% fat, 54% carbohydrates, food quotient [FQ] of 0.89) and 125 provided water ad libitum from a well-sealed water bottled. Body weight (g) was collected for 126 each animal at the start of each experiment. All mice were non-reproductive adults and were 127 housed alone, with bedding, and acclimated to experimental cages and conditions for 24-hours 128 prior to metabolic measurement. Environmental conditions for each treatment illustrated in Fig. 129 1. In brief, the desert simulation chamber has a photoperiodic cycle of 16 hr of light and 8 hrs of 130 darkness, with photoperiod transitions occurring at 06:00 (dark to light) and 20:00 (light to dark) 131 to mimic sunrise and sunset. Photoperiod remained constant across all treatments. To mirror 132 normal (baseline) circadian variation of Southwestern deserts, the environmental chamber was 133 held at a daytime temperature of 32°C and relative humidity (RH) of 10% from 09:00 to 20:00. 134 Chamber temperature and humidity were continuously recorded through the Sable Systems 135 International (SSI; Las Vegas, NV, USA) Field Metabolic System (FMS). At 20:00, the chamber

4 bioRxiv preprint doi: https://doi.org/10.1101/2020.12.18.423523; this version posted December 21, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

136 temperature is decreased and humidity increased gradually over the course of one hour to 24°C 137 and 25% RH, respectively. This one hour transition period (20:00-21:00) is hereafter referred to 138 as the evening transition (T2), whereby the room shifts from hot-and-dry to cool-and-wet 139 following lights-out at 20:00. After 21:00, conditions were held constant overnight until 06:00. 140 This period is hereafter referred to as nighttime. At 06:00, following sunrise (e.g., lights on), 141 there is a gradual 3-hour transition from cool-and-wet nighttime conditions (24°C and 25% RH) 142 to hot-and-dry daytime conditions (32°C and 10% RH), during what is hereafter referred to as 143 the morning transition period (T1). We tested two constant temperature treatments: the hot 144 treatment maintained daytime temperature and humidity (32°C and RH of 10%) levels 145 throughout the entire experiment and the constant cold treatment maintained nighttime 146 environmental conditions (24°C and 25% RH). 147 148 Metabolic phenotyping 149 Metabolic variables were measured using an 8-cage FMS manufactured by SSI. We recorded

150 oxygen consumption (VO2), carbon dioxide production (VCO2), and water vapor pressure (WVP, 151 kPa) over 72-hours for an empty baseline chamber (e.g., ambient air) and seven animal 152 chambers with an equivalent volume of 9.5 L. Air was pulled from chambers at a mass flow rate 153 of 1600 ml/min (96 L/hr), multiplexed through the SSI MUXSCAN multiplexer, and sub-sampled 154 at 250 ml/min, per the SSI FMS manual. A flow rate of 1600 ml/min and chamber volume of 9.5 155 L equate to a time constant of 5.9 minutes to sample and replace the entire container volume 156 (Lighton, 2017). Chamber sampling alternated between a baseline chamber measurement for 157 120 s followed by a random animal chamber measurement for 120 s each, resulting in 158 approximately two measurements per animal per hour. Frequent baseline measurements were 159 important for reducing measurement noise introduced by environmental transitions. This 160 protocol was repeated twice for each sex under each treatment (baseline/diurnal, hot, cold), 161 resulting in three days of continuous metabolic measurements for 14 adult males and 14 adult 162 females for each treatment. Barometric pressure (BP, Torr), temperature (degrees Celsius [°C]), 163 and relative humidity (RH, %) of the environmental chamber were recorded throughout. Raw 164 data were processed using two macros (available on Dryad) in the ExpeData analytical software -1 165 (SSI). The first macro step calculated VO2, VCO2, and VH2O (ml min at standard temperature

166 and pressure [STP]) with smoothing, z-transformation, lag correction (O2/CO2 = 18 s lag, H2O = 167 12 s lag), and gas measurements and flow rates adjusted for BP and WVP (humidity). The 168 second macro step averaged chamber measurements across the most stable 50% of each 120 169 s measurement window to produce a single average value for each metric within each window.

5 bioRxiv preprint doi: https://doi.org/10.1101/2020.12.18.423523; this version posted December 21, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

170 VO2 was used as a proxy for metabolic rate (MR), per Hulbert and Else (2004). As a measure of 171 metabolic fuel use, the respiratory quotient (RQ, L/hr), also called the respiratory exchange rate

172 (RER), was calculated as the ratio of VCO2 to VO2 (Lighton, 2017). Energy expenditure (EE, 173 kcal/hr), or the sum of the basal metabolic rate (BMR), physical activity (typically, 4-16%; Abreu- 174 Vieira et al., 2015), and cold-induced thermogenesis, was calculated as in Lighton (2017, eq. 175 9.15). BMR is difficult to measure in mice due to the confounding effects of activity, constant flux 176 in postabsorptive state, and variations in body temperature, which were not directly measured in 177 this study (Abreu-Vieira et al., 2015). Measurements located more than three standard 178 deviations away from the mean for each response variable were identified as outliers, likely non- 179 biological in nature, and were removed from downstream analyses. 180 181 Statistical Analysis 182 All analyses were conducted in R v 3.6.1 (R Core Team, 2012), unless otherwise specified. All 183 code is freely available at: www.github.com/jpcolella/peer_respo. After first confirming that the 184 data were normally distributed (shapiro_test function in the package rstatix v. 0.6.0, 185 Kassambara, 2020) and had homogeneous variance (stats::var.test), we ran an unpaired two- 186 sample t-test (stats::t.test, p < 0.05) to test for sexual dimorphism in weight. Data were subset 187 by sex, experiment, and time interval (morning transition, daytime, evening transition, nighttime) 188 to generate summary statistics. We calculated the mean (+/- sd), median, and range (minimum, 189 maximum) for each time interval, experiment, and sex. We used a student’s two-tailed t-test 190 (t.test), adjusted for multiple comparisons (Bonferroni correction, p.adjust), to test for significant 191 (p < 0.05) differences in response variables between the sexes under each experimental 192 condition. Significant differences in male and female responses (Fig. S1 and S2) led us to 193 analyze each sex independently in all downstream analyses. 194 To determine if weight was a covariate, we ran a general linear regression between 195 weight and each response variable to test for a significant relationship (R², p < 0.05). We found 196 that weight was not a significant predictor of metabolic response and thus did not include it as a 197 covariate (Fig. S3). We evaluated ANOVA assumptions in R, including: independent 198 observations, no significant outliers, data normality, and homogeneity of variance. Due to 199 ANOVA assumption violations, we elected to run a Kruskal-Wallis test (stats::kruskal.test) as a 200 non-parametric alternative to a one-way ANOVA (Hollander and Wolfe, 1973). The Kruskal- 201 Wallis test extends the two-sample Wilcoxon test to situations where there are more than two 202 groups, to identify significant differences in group means across experimental treatments (p < 203 0.05, Bonferroni corrected).

6 bioRxiv preprint doi: https://doi.org/10.1101/2020.12.18.423523; this version posted December 21, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

204 205 Time-series analyses 206 Time-series data were visualized in ggplot2 v. 3.3.0 (Wickham, 2016) and visreg v. 2.7.0 207 (Breheny and Burchett, 2017) across a 24-hour period using both local regressions, as a non- 208 parametric approach that fits multiple regressions in a local neighborhood (loess option in 209 ggplot2), and a generalized additive model (GAM; Wood, 2011) in the mgcv package v. 1.8-31 210 (Wood 2004, 2017; formula: response variable ~ s(time), where time is measured in seconds 211 across a 24-hour cycle). Plots were generated in R and edited in the free, open-source vector 212 graphics editor InkScape (https://inkscape.org). Distributions were contrasted across treatment 213 groups. 214 Change point analyses were used detect significant transitions in the mean and variance 215 (changepoint::cpt.meanvar) of each response variable across a 24-hour cycle (changepoint 216 package v. 2.2.2, Killick and Eckley, 2014; Killick et al., 2016). Change points identified under 217 baseline conditions were compared to those detected under constant hot and cold treatment 218 groups to determine whether organismal responses were anticipatory or reactive relative to 219 environmental change and whether these shifts remained unchanged or not across 220 experimental conditions. Because additional parameters almost always improve change point 221 model fit, it is important to quantitatively identify meaningful penalty thresholds to avoid 222 overparameterization (Killick and Eckley, 2014; Haynes et al., 2017). Selection of penalty 223 thresholds depends on many factors, including the size and duration of change, which remain 224 unknown prior to change point analysis (Guyon and Yao, 1999; Lavielle, 2005; Birge and 225 Massart, 2007). Current best practices are based on visual inspection of change point estimates 226 compared to time series data and known forcing functions, if available (Killick and Eckley, 227 2014). For example, there are two known (artificially controlled) environmental transitions under 228 the baseline or diurnally variable treatment conditions; therefore, a minimum of two change 229 points, maximum of four (corresponding to the beginning and end of each transition), are 230 expected and additional or fewer identified change points should be interpreted with caution and 231 explored in greater detail (Killick and Eckley, 2014). Sharp transitions in slope for example are 232 especially difficult to fit into discrete bins, and may cause additional change points to be 233 estimated within steep environmental transitions that may not be biologically relevant 234 (Fearnhead et al., 2019). To estimate penalty thresholds, we used the CROPS (Change points 235 for a Range Of Penalties; Haynes et al., 2017) penalty setting with the PELT (Pruned Exact 236 Linear Time; Killick et al., 2012) method to test a range of penalty values (λ, 5-500) and to 237 optimize computational efficiency. To balance improved model fit with the addition of more

7 bioRxiv preprint doi: https://doi.org/10.1101/2020.12.18.423523; this version posted December 21, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

238 parameters, we plotted the difference in the test statistic versus the number of change points 239 detected to identify an optimal penalty threshold of 20 as appropriate for all but one comparison 240 (e.g., Δ log-likelihood > λ). Diagnostic plots for baseline male water loss did not asymptote with 241 the x-axis under a penalty threshold of 20, thus we used a penalty value of 18 for this test. 242 Under the assumption that a large improvement in model fit is indicative of a true change point 243 (Lavielle, 2005; Killick, 2017), we used Akaike information criterion (AIC) diagnostic plots to 244 identify the optimal number of change points based on our data (maximum-likelihood). Visual 245 inspection of changepoints plotted over our raw data allowed us to identify change points that 246 may be over fit. As a Bayesian alternative, we also used the bcp package v. 4.0.3 (Erdman and 247 Emerson, 2007) to estimate change points based on a posterior probability threshold of 0.2. The 248 occurrence time of significant change points were compared across treatment groups and 249 against known room transitions to estimate lag-times and identify anticipatory versus reactionary 250 metabolic responses. The mean and slope of each segment was calculated in R. 251 252 RESULTS 253 Metabolic phenotyping 254 In total, we recorded 13,020 120-s intervals (6,550 female, 6,470 male). We measured 4,020 255 baseline observations (2,038 female and 1,982 male). We measured metabolic variables for 256 4,506 distinct 120-s intervals under the constant cold treatment: 2,256 female and 2,250 male 257 measurements. For the constant hot treatment, we measured 4,494 total observations (2,256 258 female, 2,238 male). After removing outliers, we retained 6,335 female measurements total: 259 1,877 measurements under baseline conditions, 2,230 under constant hot conditions, and 2,228 260 under constant cold conditions. We retained 6,255 measurements for males: 1,844 baseline, 261 2,214 hot, and 2,197 cold. Female weights ranged from 15.4 to 25.2 g and male weights ranged 262 from 15.5 to 26.6 g. All raw data (Expedata files) and processed machine-readable csv files are 263 available on Dryad: https://doi.org/10.5061/dryad.f4qrfj6v0. 264 265 Static statistics 266 We found no significant difference between mean weights for males and females, and no 267 relationship between weight and any measured metabolic variable (Fig. S3). T-tests between 268 the sexes identified significant differences in metabolic responses, with few exceptions (Table 1; 269 Fig. 2; Figs. S1 and S2). Summary statistics for all metabolic response variable are reported in 270 Table 1 for each sex and experimental group (baseline, hot, cold). Mean EE was higher for 271 males under all conditions, while mean water loss was only higher for males under the cold

8 bioRxiv preprint doi: https://doi.org/10.1101/2020.12.18.423523; this version posted December 21, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

272 treatment. Under diurnal conditions, RQ patterning was similar among the sexes with both 273 males and females reaching an RQ greater than 1 around midday and RQ was less than the FQ 274 overnight (Table 1). Under constant cold conditions both sexes exhibit a fixed RQ just above the 275 FQ, while under constant hot conditions female RQ was fixed and equal to the FQ, while male 276 RQ was slightly above the FQ (0.94; Table 1) but also fixed. 277 Within each sex, the majority of pairwise comparisons of metabolic variables across 278 treatment groups were significantly different (Fig. 2; Fig. S4). Separating each treatment into 279 daytime and nighttime intervals found similar evidence of significantly different physiological 280 responses among treatment groups, with few exceptions (Fig. 2; Fig. S5). Baseline daytime 281 conditions were equivalent to the hot experimental conditions, yet metabolic responses

282 unexpectedly differed between the two treatments for EE, RQ, and H2O (Table 2). As expected,

283 male VO2 did not differ between cold and baseline nighttime treatments, during which the 284 environmental conditions were identical, but these two time intervals did differed significantly in 285 all other metabolic variables examined (Fig. S2). Both sexes exhibited their highest mean EE 286 under constant cold conditions (Female = 0.23 J/s, Male = 0.25 J/s)and lowest under constant 287 hot conditions (both = 0.13 J/s). Females lost less water under diurnally variable environmental 288 conditions (mean = 0.12 mg), than they did under constant hot (0.19 mg) or cold (0.13 mg) 289 treatments, while males lost less water under the constant cold treatment (0.15 mg), relative to 290 baseline (0.16 mg) or hot (0.17 mg) treatments. 291 Weight was not correlated with any response variable (Fig. S3), therefore we did not 292 include it as a covariate or run an ANCOVA. Our data violated assumptions of normality and 293 homogeneity of variance required for an ANOVA (Tables S6, S7), so we ran nonparametric 294 Kruskal-Wallis and Wilcox tests which identified significant differences among all but five pairs 295 of group means (Fig. 2; Fig. S8). Female water loss did not differ significantly between the 296 baseline and cold treatment groups; however, differences were detected when the data were 297 split into nighttime and daytime intervals. Male EE did not differ between nighttime baseline and 298 nighttime cold treatments (Fig. 2). The most water loss was observed under the hot treatment; 299 however, peak water loss (0.43 mg for males, 0.37 mg for females) occurred under diurnally 300 variable conditions. Mean EE was higher during the nighttime, relative to daytime, for all 301 experiments. 302 303 Signal processing of physiological time-series

304 Time-series data for EE, RQ, and H2O are visualized across a 24-hour cycle in Fig. 3 for all

305 experiments and both sexes. Equivalent plots for VO2 and VCO2 are available in Fig. S9. Both

9 bioRxiv preprint doi: https://doi.org/10.1101/2020.12.18.423523; this version posted December 21, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

306 sexes exhibited similar physiological patterning of EE across all treatment groups, but at 307 different magnitudes of change (Fig. 3; Table 1). RQ was highly variable under diurnal 308 environmental cycling, but flat-lined at the FQ (0.89) under both constant conditions (Fig. 3, 4). 309 An RQ of one indicates exclusive use of carbohydrates, whereas an RQ of 0.8 indicates 310 primarily protein consumption, and an RQ of 0.7 suggests the use of fats (Richardson, 1929). 311 Under constant temperatures, RQ was fixed at 0.91 (+/- 0.07) for females and 0.90 (+/- 0.07) for 312 males under hot conditions, and at 0.88 (+/- 0.07) and 0.94 (+/- 0.10) for females and males 313 respectively, under constant cold conditions. Only under natural diurnally cycling, did RQ reach 314 values greater than one, averaging 1.00 for males and 1.05 for females during the day and 315 reaching a max of 1.37 and 1.50 for males and females, respectively. Patterns of water loss 316 were inverted between baseline and constant treatments (Fig. 3). Under baseline conditions, 317 water loss ramps up in the morning, decreases throughout the day and stabilizes at night. In 318 contrast, under constant conditions, water loss mirrors EE patterning, decreasing during the day 319 and increasing at night when animals resume higher levels of activity. Similar patterns are also

320 observed in VO2 and VCO2 (Fig. 3). 321 322 AIC optimized, maximum-likelihood change point estimates are in Table 2 and visualized in Fig.

323 4 (results for VO2 and VCO2 are visualized in Fig. S10). Under circadian environmental cycling, 324 males exhibited an average of five change points and females three. As expected for change 325 point analyses (e.g., fitting continuous data into a stepwise model), both sexes had change 326 points overfit within rapid environmental transition periods (Fig. 4). Detected metabolic shifts 327 roughly correspond to the initiation and conclusion of physical environmental transitions 328 (morning, evening). Some additional change points were detected outside of environmental 329 transition periods, including a large midday downshift in water loss present in both sexes.

330 Females exhibit a morning shift in EE, VCO2, and VO2 earlier (05:16, on a 24-hour clock 331 HH:MM) than males (05:41) on average. Morning change points for both sexes occurred prior to 332 physical changes in temperature and humidity within the environment. Females experienced a 333 second metabolic shift around 07:00 and males, just after 08:00. Evening metabolic shifts also

334 occurred in two stages. Females experience an uptick in EE, VCO2, and O2 just after 18:30, and 335 males just before 19:00, with a secondary shift around 20:12 for both sexes. We observed a 336 single, large upward morning shift in water loss during the morning environmental transition. 337 That initial uptick is followed by at least one midday downshift in water loss, followed by an 338 additional downshift coincident with (20:00 females) or soon after (20:23 males) the initiation of 339 the evening transition. Females exhibited only two change points in RQ, one during the morning

10 bioRxiv preprint doi: https://doi.org/10.1101/2020.12.18.423523; this version posted December 21, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

340 transition, after the change in photoperiod and environment, and another preemptive of the 341 evening transition. Males exhibited five AIC-optimal change points in RQ: one just before and 342 one during each transition (morning, evening) and another midday coincident with the midday 343 downshift in water loss. Means and standard deviations for each segment (i.e., period of mean 344 and variance stability between estimated change points) are reported in Table S11 for both the 345 optimal number of change points and excluding potentially overfit change points. Bayesian 346 change point results (Table S12) are generally concordant with maximum-likelihood estimates. 347 Under constant hot conditions, both sexes experienced delayed (>2 hr on average) 348 metabolic shifts relative to baseline and all major shifts occurred after changes in photoperiod. 349 Under constant cold conditions, however, the morning transition occurred later than it had under 350 baseline conditions for both sexes, but the shift remained preemptive of photoperiod change in 351 males. Evening metabolic transitions under cold conditions also moved forward in time, but 352 under this condition, the female shift occurred prior a change in photoperiod, while the male shift 353 occurred after. 354 355 DISCUSSION 356 In this study, we compared the metabolic responses of male and female cactus mice under 357 diurnally variable and constant (hot and dry, cold and wet) conditions to assess the influence of 358 temperature, humidity, and photoperiod on adaptive metabolic regulation. Overall, metabolic 359 patterning was significantly shaped by environmental conditions, females were less heat 360 tolerant than males, and metabolic responses were best optimized under diurnally variable 361 conditions. An RQ greater than one, suggestive of lipogenesis, is observed but only under 362 circadian environmental conditions, supporting the hypothesis that metabolism is well-tuned to 363 environmental cycling. A number of these observations would not have been observed under 364 constant environmental conditions and highlight the importance of continuous metabolic 365 phenotyping through environmental transitions to better understand the physiological responses 366 of organisms to dynamic environments. 367 368 Metabolic sexual dimorphism 369 Although previous studies have found different results for sexual dimorphism in this species 370 (females slightly outweigh males, Davis, 1966; no dimorphism, Manning et al., 2006), we did not 371 detect body weight dimorphism between the sexes, nor a correlation between weight and any 372 measured metabolic variable. Despite the lack of dimorphism, metabolic responses varied 373 significantly between non-reproductive adult males and females (Figs. S1, S2). While less

11 bioRxiv preprint doi: https://doi.org/10.1101/2020.12.18.423523; this version posted December 21, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

374 expected in the absence of size differences, metabolic demands frequently vary between sexes 375 (Mittendorfer, 2005; McPherson and Chenoweth, 2012; Carmona-Alcocer et al., 2012) and may 376 be explained by differences in behavior or activity level, for which EE represents a coarse proxy. 377 In general, females spent less energy than males (Table 1; Fig. 2), resulting in corresponding

378 differences in VO2, VCO2, and H2O, among other variables. Nonetheless, the overall patterns of 379 variation, including the direction, amplitude, and frequency of change across metabolic 380 response variables were similar among the sexes for all treatment groups (Fig. 3, 4). 381 Under normal diurnal cycling, respiratory water loss varied most between the sexes, with 382 males losing significantly more water than females. This observation may be in part explained 383 by higher activity levels in males (observed by not quantitatively reported herein), as increased 384 activity levels lead to greater water loss through increased respiration (Bruck, 1962). Females 385 used less energy than males during the daytime under environmentally variable conditions; Yet, 386 under constant hot conditions, females lost more daytime water than males, despite similar 387 levels of EE and identical environmental conditions across daytime diurnal and hot treatments 388 (Table 1). This suggests that females may be less tolerant of extended heat stress. The 389 opposite pattern is observed for males which appear to be less cold tolerant. Males spend 390 significantly more energy relative to females under constant cold conditions (Table 1), likely as a 391 consequence of increased thermogenesis. Consistent with this hypothesis, RQ is also lower for 392 females under cold conditions, but higher under constant hot conditions (Table 1). Even though 393 only non-reproductive individuals were examined in this study, differences in reproductive 394 biology or sex-specific selection may have influenced adaptive evolution in this species. Sexual 395 dimorphism in heat tolerance is consistent with observations in other mammals, including 396 humans (Glucksman, 1974; Mittendorfer, 2005; Pomatto et al., 2018). While endothermy 397 provides the advantages of a stable internal body temperature, which may differentially impact 398 the fitness of males and females. For example, in males, a body temperature increase of only a 399 few degrees centigrade above normothermia can reduce the viability of stored spermatozoa 400 (Moore, 1926). This challenge is typically resolved through externalizing the testes to the 401 scrotum, to maintain a lower local temperature. Although the non-reproductive cactus mice 402 examined here hold their testes inside their abdominal cavity, like most other non-reproductive 403 rodents. Even in cases of extreme heat, hyperthermia usually results in reduced fertility, albeit 404 only temporarily, as spermatozoa are quickly replaced. Yet, the fitness consequences of 405 hyperthermia may be greater in females, where the embryo cannot be thermally isolated and 406 damage is irreversible (Greenwood and Wheeler, 1985). In consequence, the actions of sex- 407 specific selection may lead to differences in traits that may not, initially, appear to be directly

12 bioRxiv preprint doi: https://doi.org/10.1101/2020.12.18.423523; this version posted December 21, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

408 associated with sexual competitiveness, such as growth rate, thermoregulation, metabolic 409 biorhythms, and environmental sensory (Glucksman, 1974; McPherson and Chenoweth, 2012; 410 Juárez-Tapia and Miranda-Anaya, 2017; Calisi et al., 2018). Difference in lipid kinetics are 411 hypothesized to maintain sexually dimorphic metabolic phenotypes (Mittendorfer, 2005). 412 Similarly genes involved in lipid metabolism have been identified as important for 413 thermoregulation in high-altitude adapted Peromyscus (Cheviron et al., 2012) and under 414 selection in wild cactus mice (Tigano et al., 2020), reinforcing an important role for sex-specific 415 lipid metabolism in the physiological adaptation of this species to desert environments. 416 417 Circadian metabolic tuning 418 We hypothesized that desert-adapted cactus mice would exhibit metabolic responses that 419 minimize water loss and limit EE under hot and dry environmental conditions. Indeed, we found 420 significant support for environmentally-associated adaptive metabolic responses for all 421 examined variables. Circadian patterning of EE is observed across all experimental treatments 422 and is consistent with the nocturnal activity pattern of this species (Warner et al., 2010; Refinetti, 423 2010): Higher at night and lower during the day. Activity patterns, like nocturnality, are known to 424 shape patterns of circadian metabolism and in mammals, shifts in activity are often governed by 425 photoperiod (Miranda-Anaya et al. 2019). In deserts, increased locomotion and food intake at 426 night minimizes evaporative water loss during hot and dry daytime conditions when 427 physiological cooling is most challenging (Abreu-Vieira et al., 2015). Animals and, more 428 specifically, near relatives of Peromyscus also maintain a warmer body temperature during the 429 active (nighttime) phase relative to the inactive (daytime) phase that cannot be wholly explained 430 by differences in activity levels, but may require additional metabolic resources to achieve 431 (Abreu-Vieira et al., 2015). EE is the sum of an organism’s BMR, physical activity, and energy 432 invested in cold-induced thermogenesis; therefore, EE represents only a coarse approximation 433 for differences in activity level. At thermoneutrality, mice (Mus) expend approximately 60% of 434 their total EE on BMR, 12% on the thermic effects of food, 25% on physical activity, and 0% on 435 cold-induced thermogenesis (Abreu-Vieira et al., 2015). Thermogenesis (Storz et al., 2019; 436 Bastias-Perez et al., 2020) likely explains both the elevated EE observed under constant cold 437 conditions and the greater variance in EE observed under diurnally variable environmental 438 conditions, where thermogenesis is expected to occur nightly but not during the day (Fig. 2; 439 Table 1). Similar patterns of EE across all treatments (high at night and lower during the day) is 440 consistent with the nocturnal activity pattern of this species and may be modulated in part by 441 photoperiod, which remained constant across all treatment groups (Miranda-Anaya et al. 2019).

13 bioRxiv preprint doi: https://doi.org/10.1101/2020.12.18.423523; this version posted December 21, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

442 The role of photoperiod in structuring activity patterns, and by proxy EE, remains to be 443 experimentally explored for this species. 444 In a captive environment where food is not a limiting resource, animals spent less 445 energy at night under constant hot conditions, than they did under diurnally variable conditions, 446 suggesting that extreme or extended temperature events may shape foraging strategies or 447 activity patterns into the future (Du Plessis et al., 2012; Mason et al., 2017; Riddell et al., 2019). 448 Behavioral thermoregulation can be achieved through increasing or decreasing activity or by 449 relocating to a thermally suitable microclimate (e.g., underground burrow). While behavioral 450 thermoregulation is energetically less expensive than autonomic thermoregulation (Terrien et 451 al., 2011; Hayford et al., 2015), there is a tradeoff between the energy and time invested in 452 thermoregulation and that which remains available for other biological processes critical to 453 survival. Increased investment in thermoregulation diverts time and energy away from 454 reproduction and resource acquisition and physical relocation to a suitable microclimate costs 455 energy and may drive animals further away from areas that are favorable in terms of resource 456 availability or predation risk (Dunbar, 1998; Kearney and Porter, 2004; Mason et al., 2017). 457 Under diurnally variable environmental conditions, both sexes experienced greater 458 variance in all metabolic response variables across a 24-hr period than under constant 459 conditions, indicating that metabolic regulation is both dynamic and environmentally mediated. 460 Cactus mice better optimize energy expenditure, water loss, and fuel resources under diurnally 461 variable environmental conditions, relative to artificially constant conditions. For example, 462 nighttime water loss was reduced under diurnal conditions, despite high activity levels, whereas 463 the reverse was true for both constant temperature treatments: high nighttime water loss and 464 EE. As expected, water loss was greatest under constant hot and dry conditions and lowest 465 under constant cold and wet conditions (Table 1). However, when split into daytime and 466 nighttime intervals, animals in the diurnal treatment group lost significantly less water (Fig. 1), 467 even when conditions were identical (e.g., daytime diurnal and hot conditions). In fact, during 468 the daytime under diurnally variable conditions, animals lost even less water than they did under 469 constant cold conditions, suggesting the water economy of cactus mice is most efficient under 470 natural diurnal cycling or when the animals are not thermally stressed. Although significant, the 471 volumetric difference in water loss between treatments is small (0.01-0.07), yet the biological 472 consequences of these differences remain to be tested in detail. 473 474 Lipogenesis as potential a buffer against dehydration

14 bioRxiv preprint doi: https://doi.org/10.1101/2020.12.18.423523; this version posted December 21, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

475 The impact of environmental variation on cactus mouse metabolism is most obvious for RQ. 476 Under diurnally variable conditions, RQ increased with increasing temperature and humidity, 477 followed by an evening downshift, coincident with or occurring just prior to the evening transition 478 to cooler, wetter conditions (Figs. 3, 4). In contrast, RQ flat-lined at the FQ under both constant 479 conditions. This indicates that RQ in cactus mice is tightly modulated by temperature, and more 480 specifically, sustained temperatures. 481 An RQ over one can indicate either activity levels above aerobic threshold (Whipp, 2007; 482 Zagatto et al., 2012) or lipogenesis (Benedict and Lee, 1937; Kleiber, 1961; Elia and Livesey, 483 1988; Abreu-Vieira et al., 2015; Levin et al., 2017). For cactus mice, elevated RQ levels are 484 observed during the daytime, when mice are inactive and are therefore suggestive of facultative 485 lipogenesis. Glycolysis converts carbohydrates (glucose) into pyruvate, which is then converted 486 into acetyl-CoA. Excess acetyl-CoA can be used to produce fatty acids and triglycerides through 487 the process of lipogenesis. During this process, glycerol hydroxyl groups (-OH) react with the

488 carboxyl end (-COOH) of a fatty acid chain to produce water (H2O) and carbon (C) atoms that 489 then bind to oxygen (O) through a dehydration synthesis reaction, forming water and carbon 490 dioxide (Ameer et al., 2014; OpenStax, 2016). In consequence, lipogenesis results in a

491 measurable uptick in released CO2, as evidenced by a spike in RQ (the ratio of VCO2 to VO2). 492 Since free water is scarce in desert environments, we hypothesize that the breakdown of food 493 via lipogenesis may be important indirect source of water for cactus mice (Bradley and Yousef, 494 1972), whereby the endogenous synthesis of fatty acids and triglycerides may buffer against the 495 negative physiological effects of dehydration. Interestingly, an RQ greater than one, consistent 496 with lipogenesis, is only observed during the daytime under diurnally variable environmental 497 conditions and surprisingly, never recorded under constant hot and dry conditions. Circadian 498 cycling of lipogenesis is observed in other rodents (hamsters, rats, mice), modulated by dietary 499 and hormonal variation (insulin and prolactin, Kimura et al., 1970; Cincotta and Meier, 1985; 500 Bryson et al., 1993; Kersten, 2001; Letexier et al., 2003; Abreu-Vieira et al., 2015). Insulin- 501 associated genes have been shown to be under selection in cactus mice (Insl3, Igfbp3; 502 Kordonowy et al., 2017) and Inpp5k appears to be a functionally important transcript for water 503 regulation in this species, that is also involved in insulin signaling (MacManes, 2017; Bridges 504 and Saltiel, 2015).

505 Circadian patterning of RQ, and therefore also VCO2 and VO2, in cactus mice under 506 environmentally variable conditions is likely what was described as ‘circadian torpor’ by 507 Macmillen (1965, 1972). While the abiotic regulatory mechanism of lipogenesis in cactus mice 508 remains elusive, it is not driven by photoperiod, nor constant temperature or humidity. Circadian

15 bioRxiv preprint doi: https://doi.org/10.1101/2020.12.18.423523; this version posted December 21, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

509 lipogenesis also does not appear to be resource-dependent, as originally hypothesized by 510 Macmillen (1965, 1972; also see Stephan, 2001), because animals were provided food and 511 water ad libitum throughout the experiments. Potentially consistent with Macmillen’s (1965, 512 1972) resource-limitation hypothesis, however, RQ patterning suggests that lipogenesis may be 513 glucose-limited, at least within a single day. The lipogenesis reaction is controlled by hepatic 514 insulin levels (Ameer et al., 2014), thus, we hypothesize that lipogenesis decreases from 515 morning to evening in response to decreasing insulin levels or leptin-ghrelin signaling (Buijs et 516 al., 2017; Luna-Moreno et al., 2017), which are not mutually exclusive. For example, leptin 517 regulates hepatic insulin sensitivity and glucose homeostasis, as well as energy expenditure 518 (Inoue et al., 2004). Insulin levels are expected to become depleted over the course of the day, 519 while nighttime refeeding is anticipated to increase serum insulin, and in turn, potentially 520 stimulate lipogenesis (Kimura et al., 1970; Bryson et al., 1993; Yamajuku et al., 2012; Liu et al., 521 2013; Adamovich et al., 2014). The absence of an RQ greater than one under either constant 522 treatment suggests that this pattern is not solely mediated by insulin availability. Extended 523 exposure to sub-optimal environmental conditions can negatively impact body condition and 524 survival in endotherms (McKechnie and Wolf, 2010; Gardner et al., 2016); therefore, we 525 hypothesize that the detection of extended sub-optimal conditions triggers a downshift in 526 metabolic resource consumption in response to physiological stress (Taborsky et al., 2020). In 527 addition to the physiological limits of an organism, environmental stability is a key factor in 528 shaping stress responses (Taborsky et al., 2020). Continuous metabolic phenotyping through 529 state transitions (multiple days of diurnal cycling to multiple days of constant conditions) will be 530 necessary to determine the abiotic triggers and metabolic processes responsible for circadian 531 variation in RQ. 532 High altitude adapted Peromyscus exhibit metabolic differences that enhance survival in 533 cold, hypoxic (Ivy and Scott, 2018, 2020) environments. In these species, cold-induced 534 thermogenesis is used to maintain a constant internal body temperature and is regulated by 535 flexibility in lipid metabolism (Cheviron et al., 2012; Munshi-South and Richardson, 2017). Lipids 536 are also known to regulate circadian rhythmicity in mammals, with hepatic lipid metabolism 537 being controlled by a suite of circadian Clock genes (Eckel-Mahan et al., 2012; Adamovich et 538 al., 2014, 2015) some of which have been identified as under selection in cactus mice (BMAL1, 539 BMAL2; Colella et al., in press). Under acute dehydration, cactus mice lose > 20% of their body 540 weight without neurological, behavioral, or cardiovascular impairment, which would normally 541 impact other mammals (Kordonowy et al., 2017). Developmental exposure to thermal extremes

16 bioRxiv preprint doi: https://doi.org/10.1101/2020.12.18.423523; this version posted December 21, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

542 can also shape adaptive responses (Ivy and Scott, 2020) and the role of development in 543 metabolic regulation of dehydration tolerance remains to be explored for this species. 544 545 Timing of metabolic shifts 546 Analyses of continuous metabolic phenotypes confirmed that cactus mice physiology is well- 547 tuned to circadian patterns of environmental variation. We expected there to be a physiological 548 lag between the sensation of environmental change and a measurable shift in the circadian 549 system (Aguilar-Roblero, 2015). For example, in related a Cricetid species, Neotomodon alstoni, 550 increased activity levels are observed immediately following the initiation of the dark phase 551 (e.g., lights out; Miranda-Anaya et al. 2019). In contrast, under normal circadian cycling, most 552 metabolic shifts in cactus mice occurred in advance of physical environmental change, or the 553 perception thereof (Fig. 3, 4). Evidence of anticipatory metabolic shifts support the hypothesis 554 that entrainment of biorhythms is adaptive in highly variable environments. 555 The timing of most metabolic shifts is disrupted under constant temperature conditions; 556 thus, circadian metabolic patterning is not exclusively regulated by access to food, temperature, 557 or humidity. The persistence of two major metabolic shifts under both constant treatments, 558 however, suggests that photoperiod may be at least in part responsible for maintaining 559 biorhythms. Photic synchronization is common among mammals, in which the hypothalamus 560 acts as a circadian pacemaker, reactive to ambient light (Miranda-Anaya et al. 2019). Observed 561 temporal changes in metabolic shifts between baseline and constant treatment groups were 562 consistent with males being more heat tolerant than females. Relative to males, females use 563 less energy and lose less water under cold conditions, but experienced a longer period of 564 nighttime activity under hot conditions. Tests of inverted day and night conditions will be 565 necessary to test the strength of the observed circadian rhythm and role of photoperiod. 566 567 Conclusions 568 In conclusion, our results highlight an important role for circadian environmental variation in the 569 evolution of metabolic efficiency in desert-adapted cactus mice. Lipid metabolism is increasingly 570 emerging as a common mechanism of circadian and thermoregulatory plasticity, that warrants 571 further experimental investigation in this species through both dietary and environmental 572 manipulation and in vitro tests of metabolic fuel usage (Adamovich et al., 2014). Studies of 573 physiological in nontraditional model organisms (Bedford and Hoekstra 2015) provides a 574 comparative framework for identifying shared physiological responses and mechanisms among 575 species. Further, linking molecular mechanisms to the physiological phenomena described here

17 bioRxiv preprint doi: https://doi.org/10.1101/2020.12.18.423523; this version posted December 21, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

576 will allow us to better characterize the role of genetics, environment, and behavior on complex 577 physiological phenotypes (Partch et al., 2014). 578 579 ACKNOWLEDGEMENTS 580 We thank all members of the MacManes Laboratory for useful comments on early versions of 581 the manuscript; the Animal Resources Office and veterinary care staff at the University of New 582 Hampshire; A. Gerson at the University of Massachusetts Amherst for analytic guidance and 583 technical support; Z. Cheviron at the University of Montana; B. Joos and J. Klok at Sable 584 Systems International for their guidance in setting up our respirometry system; and A. 585 Westbrook for computational support. This work was supported by the National Institute of 586 Health National Institute of General Medical Sciences [1R35GM128843 to M.D.M.]. 587 588 COMPETING INTERESTS 589 No competing interests declared. 590 591 DATA AVAILABILITY 592 Raw ExpeData (SSI) files are available through DataDryad: 593 https://doi.org/10.5061/dryad.f4qrfj6v0. Macro processing files, respirometry data, and cage 594 sampling scheme files are available also on Data Dryad. All R scripts used in this project are 595 available through GitHub at: github.com/jpcolella/peer_respo. 596 597 REFERENCES 598 Abreu-Vieira, G., Xiao, C., Gavrilova, O., and Reitman, M. L. (2015). Integration of body 599 temperature into the analysis of energy expenditure in the mouse. Mol. Metab., 4, 461-470. 600 Adamovich, Y. Arviram, R., and Asher, G. (2015). The emerging roles of lipids in circadian 601 control. Biochem. Biophys. Acta., 1851, 1017-1025. 602 Adamovich, Y., Rousso-Noori, L., Zwighaft, Z., Neufeld-Cohen, A., Golik, M., Kraut-Cohen, 603 J., Wang, M., Han, X., Asher, G. (2014). Circadian clocks and feeding time regulate oscillations 604 and levels of hepatic triglycerides. Cell Metab., 19, 319-330. 605 Aguilar-Roblero, R. (2015). Introduction to circadian rhythms, clocks and its genes. In: 606 Mechanisms of circadian systems in aniamls and their clinical relevance (Eds. R. A., Robler, M. 607 Díaz-Muñoz, M. L., Fanjul-Moles). Springer Cham, Heidelberg New York Dordrecht London. Pp. 608 1-14.

18 bioRxiv preprint doi: https://doi.org/10.1101/2020.12.18.423523; this version posted December 21, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

609 Alonso, J., Salgado, I., and Palacín, C. (2016). Thermal tolerance may cause sexual 610 segregation in sexually dimorphic species living in hot environments. Behav. Ecol., 27, 717– 611 724. 612 Ameer, F., Scandiuzzi, L., Hasnain, S., Kalbacher, H., and Zaidi, N. (2014). De novo 613 lipogenesis in health and disease. Metabolism, 63, 895–902. 614 Bastías-Pérez, M., Zagmutt, S., Soler-Vázquez, M., Serra, D., Mera, P., and Herrero, L. 615 (2020). Impact of adaptive thermogenesis in mice on the treatment of obesity. Cells, 9, 316. 616 Bedford, N. L., and Hoekstra, H. E. (2015). The natural history of model organisms: 617 Peromyscus mice as a model for studying natural variation. Elife, 4, e06813. 618 Benedict, F., and Lee, R. (1937). Lipogenesis in the animal body, with special reference to the 619 physiology of the goose. Carnegie Inst. Washington Pub. 620 Bennett, A., and Ruben, J. (1979). Endothermy and activity in vertebrates. Science, 206, 649– 621 654. 622 Birge, L., and Massart, P. (2007). Minimal penalties for Gaussian model selection. Prob. 623 Theory Rel., 138, 33–73. 624 Boyles, J., Seebacher, F., Smit, B., and McKechnie, A. (2011). Adaptive thermoregulation in 625 endotherms may alter responses to climate change. Integr. Comp. Biol., 51, 676–690. 626 Bradley, W., and Yousef, M. (1972). Small mammals in the desert. In Physiological 627 adaptations: Desert and mountain (pp. 127–142). New York: Academic Press. 628 Breheny, P., and Burchett, W. (2017). Visualization of regression models using visreg. The R 629 Journal, 9, 56–71. 630 Bridges, D., and Saltiel, A. (2015). Phosphoinositides: Key modulators of energy metabolism. 631 Biochim. Biophys. Acta, 1851, 857–866. 632 Bruck, E. (1962). Water in expired air: Physiology and measurement. J. Pediatrics, 60, 869– 633 881. 634 Bryson, J., Cooney, G., Wensley, V., Blair, S., and Caterson, I. (1993). Diurnal patterns of 635 cardiac and hepatic pyruvate dehydrogenase complex activity in gold-thioglucose-obese mice. 636 Biochem. J., 295, 731–734. 637 Buijs, F. N., Guzmán-Ruiz, M., León-Mercado, L., Basualdo, M. C., Escobar, C., Kalsbeek, 638 A., and Buijs, R. M. (2017). Suprachiasmatic nuclear interaction with the arcuate nuclear; 639 essential for organizing physiological rhythms. eNeuro, 4, e0028-17.2017. 640 Calisi, R., Austin, S., Lang, A., and MacManes, M. (2018). Sex-biased transcriptomic 641 response of the reproductive axis to stress. Horm. Behav., 100, 56–68.

19 bioRxiv preprint doi: https://doi.org/10.1101/2020.12.18.423523; this version posted December 21, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

642 Carmona-Alcocer, V., Fuentes-Granados, C., Carmona-Castro, A., Aguilar-González, 643 Cárdenas-Vázquez, R., and Miranda-Anaya, M. (2012). Obesity alters circadian behavior and 644 metabolism in sex dependent manner in the volcano mouse Neotomodon alstoni. Physiol. 645 Behav., 105, 727-733. 646 Cheuvont, S., Kenefick, R., Montain, S., and Sawka, M. (2010). Mechanisms of aerobic 647 performance impairment with heat stress and dehydration. J. Appl. Physiol., 109, 1989–1995. 648 Cheviron, Z., Bachman, G., Connaty, A., McClelland, G., and Storz, J. (2012). Regulatory 649 changes contribute to the adaptive enhancement of thermogenic capacity in high-altitude deer 650 mice. P. Natl. A. Sci. USA, 109, 8635–8640. 651 Cheviron, Z., Connaty, A., McClelland, G., and Storz, J. (2014). Functional genomics of 652 adaptation to hypoxic cold-stress in high-altitude deer mice: Transcriptomic plasticity and 653 thermogenic performance. Evolution, 68, 48–62. 654 Cincotta, A., and Meier, A. (1985). Prolactin permits the expression of a circadian variation in 655 lipogenesis responsiveness to insulin in hepatocytes of the golden hamster (Mesocricetus 656 auratus). J. Endocrinol., 106, 173–176. 657 Colella, J. P., Tigano, A., Dudchenko, O., Omer, A. D., Khan, R., Bochkov, I. D., Aiden, E. 658 L., MacManes, M. D. (in press). Limited evidence for parallel evolution among desert-adapted 659 Peromyscus deer mice. J. Heredity, in press. 660 Dalziel, A., Rogers, S., and Schulte, P. (2009). Linking genotypes to phenotypes and fitness: 661 How mechanistic biology can inform molecular ecology. Molecular Ecology, 18, 4997–5017. 662 Davis, W. (1966). The mammals of Texas. Bulletin of the Texas Game and Fish Commission, 663 41, 5–267. 664 Degen, A. (2012). Ecophysiology of small desert mammals. Springer Science and Business 665 Media. 666 Dunbar, R. (1998). Impact of global warming on the distribution and survival of the gelada 667 baboon: A modelling approach. Global Change Biol., 4, 293–304. 668 du Plessis, K., Martin, R., Hockey, P., Cunningham, S., and Ridley, A. (2012). The costs of 669 keeping cool in a warming world: Implications of high temperatures for foraging, 670 thermoregulation and body condition of an arid-zone bird. Global Change Biol., 18, 3063–6070. 671 Easterling, D., Meehl, G., Parmesan, C., Changnon, S., Karl, T., and Mearns, L. (2000). 672 Climate extremes: Observations, modeling, and impacts. Science, 289, 2068–2074. 673 Eckel-Mahan, K. L., Patel, V. R., Mohney, R. P., Vignola, K. S., Baldi, P., and Sassone- 674 Corsi, P. (2012). Coordination of the transcriptome and metabolome by the circadian clock. 675 Proc. Natl. Acad. Sci. USA, 109, 5541-5546.

20 bioRxiv preprint doi: https://doi.org/10.1101/2020.12.18.423523; this version posted December 21, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

676 Elia, M., and Livesey, G. (1988). Theory and validity of indirect calorimetry during net lipid 677 synthesis. Am. J. Clin. Nutr., 47, 591–607. 678 Erdman, C., and Emerson, J. (2007). bcp: An R package for performing a Bayesian analysis of 679 change point problems. J. Stat. Softw., 23, 1–13. 680 Fearnhead, P., Maidstone, R., and Letchford, A. (2019). Detecting changes in slope with an 681 L0 penalty. J. Comput. Graph. Stat., 28, 265–275. 682 Fuglei, E., and Oritsland, N. (1999). Seasonal trends in body mass, food intake and resting 683 metabolic rate, and induction of metabolic depression in arctic foxes (Alopex lagopus) at 684 Svalbard. J. Comp. Phys., 169, 361–369. 685 Gardner, J., Amano, T., Sutherland, W., Clayton, M., and Peters, A. (2016). Individual and 686 demographic consequences of reduced body condition following repeated exposure to high 687 temperatures. Ecology, 97, 786–795. 688 Garland, T., and Carter, P. (1994). Evolutionary physiology. Annu. Rev. Physiol., 56, 579–621. 689 Gienapp, P., Teplitsky, C., Alho, J., Mills, J., and Merilä, J. (2008). Climate change and 690 evolution: Disentangling environmental and genetic responses. Mol. Ecol., 17, 167–178. 691 Glucksmann, A. (1974). Sexual dimorphism in mammals. Biol. Rev., 49, 423–475. 692 Greenwood, P., and Wheeler, P. (1985). The evolution of sexual size dimorphism in birds and 693 mammals: A “hot blooded” hypothesis. In Evolution: Essays in honour of John Maynard Smith 694 (Greenwood, P. J., Harvey, P. H., Slatkin M. (Eds.), pp. 287–299). New York: Cambridge 695 University Press. 696 Guyon, X., and Yao, J. (1999). On the underfitting and overfitting sets of models chosen by 697 order selection criteria. J. Multivariate Anal., 70, 221–249. 698 Harris, S., and Munshi-South, J. (2017). Signatures of positive selection and local adaptation 699 to urbanization in white-footed mice (Peromyscus leucopus). Mol. Ecol., 26, 6336–6350. 700 Haussinger, D. (1996). The role of cellular hydration in the regulation of cell function. Biochem. 701 J., 313, 697–710. 702 Hayford, H., Gilman, S., and Carrington, E. (2015). Foraging behavior minimizes heat 703 exposure in a complex thermal landscape. Mar. Ecol. Prog. Ser., 518, 165–175. 704 Heimeier, R., Davis, B., and Donald, J. (2002). The effect of water deprivation on the 705 expression of atrial natriuretic peptide and its receptors in the spinifex hopping mouse Notomys 706 alexis. Comp. Biochem. Phys. A, 132, 893–903. 707 Hock, R. J. (1964). Physiological responses of deer mice to various native altitudes. In: The 708 Physiological effects of High Altitude (ed. Weihe, W.), pp. 59-72. New York, NY: Macmillan.

21 bioRxiv preprint doi: https://doi.org/10.1101/2020.12.18.423523; this version posted December 21, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

709 Hollander, M., and Wolfe, D. (1973). Nonparametric Statistical Methods. New York: John Wiley 710 and Sons. 711 Huey, R., Kearney, M., Krockenberger, A., Holtum, J., Jess, M., and Williams, S. (2012). 712 Predicting organismal vulnerability to climate warming: Roles of behaviour, physiology and 713 adaptation. Philos. T. R. Soc. B., 367, 1665–1679. 714 Hulbert, A., and Else, P. (2004). Basal metabolic rate: History, composition, regulation, and 715 usefulness. Physiol. Biochem. Zool, 77, 869–876. 716 IPCC. (2012). Managing the risks of extreme events and disasters to advance climate change 717 adaptation: Special report of the Intergovernmental Panel on Climate Change (Field, C.B., V. 718 Barros, T.F. Stocker, D. Qin, D.J. Dokken, K.L. Ebi, M.D. Mastrandrea, K.J. Mach, G.-K. 719 Plattner, S.K. Allen, M. Tignor, and P.M. Midgley). Cambridge University Press. 720 IPCC. (2018). Global Warming of 1.5° C: An IPCC special report on the impacts of global 721 warming of 1.5° C above pre-industrial levels and related global greenhouse gas emission 722 pathways, in the context of strengthening the global response to the threat of climate change, 723 sustainable development, and efforts to eradicate poverty. Intergovernmental Panel on Climate 724 Change. 725 Ivy, C. M., and Scott, G. R. (2018). Evolved changes in breathing and CO2 sensitivity in deer 726 mice native to high altitudes. Am. J. Physiol. Integr. Comp. Physiol., 315, R1027-1037. 727 Ivy, C. M., and Scott, G. R. (2020) Life-long exposure to hypoxia affects metabolism and 728 respiratory physiology across life stages in high-latitude deer mice (Peromyscus maniculatus). 729 J. Exp. Biol., in press. doi: 10.1242/jeb.237024. 730 Juárez-Tapia, C., and Miranda-Anaya, M. (2017). Ovariectomy influences the circadian rhythm 731 of locomotor activity and the photic phase shifts in the volcano mouse. Physiol. Behav., 182, 77- 732 85. 733 Kassambara, A. (2020). rstatix: Pipe-friendly framework for basic statistical tests (Version 734 0.6.0) [R v. 3.6.1]. Retrieved from https://CRAN.R-project.org/package=rstatix 735 Haynes, K., Eckley, I., and Fearnhead, P. (2017). Computationally efficient changepoint 736 detection for a range of penalties. J. Comp. Graph. Stat., 26, 134–143. 737 Kearney, M., Shine, R., and Porter, W. (2009). The potential for behavioral thermoregulation to 738 buffer “cold-blooded” animals against climate warming. P. Natl. Acad. Sci. USA, 106, 3835– 739 3840. 740 Kearney, M., and Porter, W. (2004). Mapping the fundamental niche: Physiology, climate, and 741 the distribution of a nocturnal lizard. Ecology, 85, 3119–3131.

22 bioRxiv preprint doi: https://doi.org/10.1101/2020.12.18.423523; this version posted December 21, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

742 Kersten, S. (2001). Mechanisms of nutritional and hormonal regulation of lipogenesis. EMBO 743 Rep., 2, 282–286. 744 Killick, R. (2017). Introduction to optimal changepoint detection algorithms. useR! Tutorial. 745 Retrieved from http://members.cbio.mines-paristech.fr/~thocking/change-tutorial/RK- 746 CptWorkshop.html#references 747 Killick, R., and Eckley, I. (2014). changepoint: An R package for changepoint analysis. J. Stat. 748 Softw., 58, 1–19. 749 Killick, R., Fearnhead, P., and Eckley, I. (2012). Optimal detection of changepoints with a 750 linear computational cost. J. Am. Stat. Assoc., 107, 1590–1598. 751 Killick, R., Haynes, K., and Eckley, I. (2016). changepoint: An R package for changepoint 752 analysis (Version 2.2.2). Retrieved from https://CRAN.R-project.org/package=changepoint 753 Kimura, T., Maji, T., and Ashida, K. (1970). Periodicity of food intake and lipogenesis in rats 754 subjected to two different feeding plans. J. Nutr., 100, 691–697. 755 Kingsolver, J. (2009). The well-temperatured biologist. Am. Nat., 174, 755–768. 756 Kleiber, M. (1961). The Fire of Life: An introduction to animal energetics. London: John Wiley 757 and Sons, Inc. Retrieved from https://www.cabdirect.org/cabdirect/abstract/19621404881 758 Kordonowy, L., Lombardo, K., Green, H., Dawson, M., Bolton, E., LaCourse, S., and 759 MacManes, M. (2016). Physiological and biochemical changes associated with acute 760 experimental dehydration in the desert adapted mouse, Peromyscus eremicus. Phys. Rep., 5, 761 e13218. 762 Lavielle, M. (2005). Using penalized contrasts for the change-point problem. Signal Process., 763 85, 1501–1510. 764 Letexier, D., Diraison, F., and Beylot, M. (2003). Addition of insulin to a moderately high- 765 carbohydrate diet reduces hepatic lipogenesis and plasma tricylglycerol concentrations in 766 humans. The American J. Clin. Nutr., 77, 559–564. 767 Levin, E., Lopez-Martinez, G., Fane, B., and Davidowitz, G. (2017). Hawkmoths use nectar 768 sugar to reduce oxidative damage from flight. Science, Signal Process., 355, 733–735. 769 Lighton, J. (2017). Limitations and requirements for measuring metabolic rates: A mini review. 770 Eur. J. Clin. Nutr., 71, 301–305. 771 Liu, S., Brown, J., Stanya, K., Homan, E., Leidl, M., Inouye, K., Bhargava, P., Gangl, M. R., 772 Dai, L., Hatano, B. et al. (2013). A diurnal serum lipid integrates hepatic lipogenesis and 773 peripheral fatty acid utilization. Nature, 502, 550–554.

23 bioRxiv preprint doi: https://doi.org/10.1101/2020.12.18.423523; this version posted December 21, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

774 Luna-Moreno, D., Pérez-Mendoza, M., Carmona-Castro, A., and Miranda-Anaya, M. (2017). 775 Daily profile in ghrelin and hypothalamic ghrelin receptors in obese and lean mice Neotomodon 776 alstoni. Biol. Rhythm Res., 49, 515-520. 777 MacManes, M. (2017). Severe acute dehydration in a desert rodent elicits a transcriptional 778 response that effectively prevents kidney injury. Am. J. Physiol.-Renal, 313, F262–F272. 779 MacManes, M., and Eisen, M. (2014). Characterization of the transcriptome, nucleotide 780 sequence polymorphism, and natural selection in the desert adapted mouse Peromyscus 781 eremicus. PeerJ, 2, e642. 782 Macmillen, R. (1965). Aestivation in the cactus mouse Peromyscus eremicus. Comp. Biochem. 783 Physiol., 16, 227–248. 784 Macmillen, R. (1972). Water economy of nocturnal desert rodents. Sym. Zool. S., 31, 147–174. 785 Manning, R., Yancey II, F., and Jones, C. (1-8). Morphometric variation in two populations of 786 the cactus mouse (Peromyscus eremicus) from trans-pecos Texas. Occas. Pap. Museum Texas 787 Tech Uni., 262, 2006. 788 Mason, T., Brivio, F., Stephens, P., Apollonio, M., and Grignolio, S. (2017). The behavioral 789 trade-off between thermoregulation and foraging in a heat-sensitive species. Behav. Ecol., 28, 790 908–918. 791 McKechnie, A., and Wolf, B. (2010). Climate change increases the likelihood of catastrophic 792 avian mortality events during extreme heat waves. Biol. Letters, 6, 253–256. 793 McNab, B., and Morrison, P. (1963). Body temperatures and metabolism in subspecies of 794 Peromyscus from arid and mesic environments. Ecol. Monogr., 33, 63–82. 795 McPherson, F., and Chenoweth, P. (2012). Mammalian sexual dimorphism. Animal Repro. 796 Sci,, 131, 109–122. 797 Merkt, J., and Taylor, R. (1994). “Metabolic switch” for desert survival. P. Natl. Acad. Sci. 798 USA., 91, 12313–12316. 799 Miranda-Anaya, M., Pérez-Mendoza, M., Juárez-Tapia, C. R., Carmona-Castro, A. (2019). 800 The volcano mouse Neotomodon alsoni of central Mexico, a biological model in the study of 801 breeding, obesity, and circadian rhythms. Gen. Comp. Endocrin., 273, 61-66. 802 Mittendorfer, B. (2005). Sexual dimorphism in human lipid metabolism. J. Nutr., 135, 681–686. 803 Montain, S., Latzka, W., and Sawka, N. (1999). Fluid replacement recommendations for 804 training in hot weather. Mil. Med., 164, 502–508. 805 Moore, C. (1926). The biology of the mammalian testis and scrotum. Q. Rev. Biol., 1, 4–50. 806 Morrison, S., Nakamura, K., and Madden, C. (2008). Central control of thermogenesis in 807 mammals. Exp. Physiol., 93, 773–797.

24 bioRxiv preprint doi: https://doi.org/10.1101/2020.12.18.423523; this version posted December 21, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

808 Munshi-South, J., and Richardson, J. (2017). Peromyscus transcriptomics: Understanding 809 adaptation and gene expression plasticity within and between species of deer mice. Semin. 810 Cell. Dev. Biol., 61, 131–139. 811 Murie, M. (1961). Metabolic characteristics of mountain, desert, and coastal populations of 812 Peromyscus. Ecology, 42, 723–740. 813 OpenStax CNX. (2016). Anatomy and Physiology. OpenStax CNX. Retrieved from 814 http://cnx.org/contents/[email protected] 815 Parmesan, C., and Yohe, G. (2003). A globally coherent fingerprint of climate change impacts 816 across natural systems. Nature, 421, 37–42. 817 Partch, C. L., Green, C. B., and Takahashi, J. S. (2014). Molecular architecture of the 818 mammalian circadian clock. Trends Cell. Biol., 24, 90-99. 819 Pomatto, L., Tower, J., and Davies, K. (2018). Sexual dimorphism and aging differentially 820 regulate adaptive homeostasis. J. Gerontol. A-Biol., 73, 141–149. 821 R Core Team. (2012). R: A language and environment for statistical computing. R Foundation 822 for Statistical Computing (Version 3.6.1). Vienna, Austria. Retrieved from http://www.R- 823 project.org/ 824 Refinetti, R. (2010). The circadian rhythm of body temperature. Front. Biosci., 15, 564-594. 825 Reid, W. (1987). A critical analysis of Eastern Mojave desert temperatures [Dissertation]. 826 Northridge: California State University. 827 Richardson, H. (1929). The respiratory quotient. Physiol. Rev., 9, 61–125. 828 Ruf, T., and Geiser, F. (2015). Daily torpor and hibernation in birds and mammals. Biol. Rev., 829 90, 891–926. 830 Seebacher, F. (2009). Responses to temperature variation: Integration of thermoregulation and 831 metabolism in vertebrates. J. Exp. Biol., 212, 2885–2891. 832 Stephan, F. K. (2001). Food entrainable oscillators in mammals. Handbook of Behavioral 833 Neurobiology Circadian Clocks, vol. 12. (Eds. J. D. Takahashi, F. W. Turek, R. Y. Moore). 834 Kluwer Acad. Press, New York, London. Pp. 223-241. 835 Storz, J., Cheviron, Z., McClelland, G., and Scott, G. (2019). Evolution of physiological 836 performance capacities and environmental adaptation: Insights from high-elevation deer mice 837 (Peromyscus maniculatus). J. Mamm., 100(3), 910–922. 838 Taborsky, B., English, S., Fawcett, T. W., Kuijper, B., Leimar, O., McNamara, J. M., 839 Ruuskanen, S., and Sandi, C. (2020). Towards an evolutionary theory of stress responses. 840 Trends in Ecol. Evol.: in press. doi: 10.1016/j.tree.2020.09.003.

25 bioRxiv preprint doi: https://doi.org/10.1101/2020.12.18.423523; this version posted December 21, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

841 Terrien, J., Parret, M., and Aujard, F. (2011). Behavioral thermoregulation in mammals: A 842 review. Front. Biosci., 16(4), 1428–1444. 843 Tigano, A., Colella, J., and MacManes, M. (2020). Comparative and population genomics 844 approaches reveal the basis of adaptation to deserts in a small rodent. Molec. Ecol., 29, 1300– 845 1314. 846 Urban, M. (2015). Accelerating extinction risk from climate change. Science, 348, 571–573. 847 Vale, G., and Brito, J. (2015). Desert-adapted species are vulnerable to climate change: 848 Insights from the warmest region on Earth. Global Ecol. Conserv., 4, 369–379. 849 Warner, A., Jethwa, P., Wyse, C., l’Anson, H., Brameld, J., and Ebling, F. (2010). Effects of 850 photoperiod on daily locomotor activity, energy expenditure, and feeding behavior in a seasonal 851 mammal. Am. J. Physiol.-Reg. I, 298, R1409–R1416.

852 Whipp, B. (2007). Physiological mechanisms dissociating pulmonary CO2 and O2 exchange 853 dynamics during exercise in humans. Exp. Physiol., 92, 347–355. 854 Wickham, H. (2016). ggplot2: Elegant graphics for data analysis. New York: Springer-Verlag. 855 Retrieved from http://ggplot2.org 856 Wood, S. (2004). Stable and efficient multiple smoothing parameter estimation for generalized 857 additive models. J. Am. Stat. Assoc., 99, 673–686. 858 Wood, S. (2011). Fast stable restricted maximum likelihood and marginal likelihood estimation 859 of semiparametric generalized linear models. J. Roy. Stat. Soc. B, 73, 3–36. 860 Wood, S. (2017). Generalized Additive Models: An Introduction with R (2nd edition). Chapman 861 and Hall/CRC. 862 Yamajuku, D., Inagaki, T., Haruuma, T., Okubo, S., Kataoka, Y., Kobayashi, S., Kegami, K., 863 Laurent, T., Kojima, T., Noutomi, K., Hashimoto, S., and Oda, H. (2012). Real-time 864 monitoring in three dimensional hepatocytes reveals that insulin acts as a synchronizer for liver 865 clock. Sci. Rep., 2, 439. 866 Zagatto, A., Miyagi, W., Sakugawa, R., Kaminagakura, E., and Papoti, M. (2012). Aerobic 867 endurance measurement by respiratory exchange ratio during a cycle ergometer graded 868 exercise test. J. Exercise. Physiol., 15, 49–56. 869 870 871 872 FIGURES AND FIGURE LEGENDS 873 Figure 1. Schematic of environmental variables across a 24-hour cycle for each treatment 874 group: diurnally variable baseline treatment in yellow, constant hot treatment in orange, and

26 bioRxiv preprint doi: https://doi.org/10.1101/2020.12.18.423523; this version posted December 21, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

875 constant cold treatment in blue. Temperature is indicated by a solid line and relative humidity (%% 876 RH) by dashed lines. Photoperiod is indicated by light and dark grey boxes. 877

878

27 bioRxiv preprint doi: https://doi.org/10.1101/2020.12.18.423523; this version posted December 21, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

Figure 2. Violin plots showing mean (black dot) for energy expenditure (EE), respiratory quotient (RQ), and water loss (H2O). Standard deviation is shown as a horizontal black line and all observations are represented by grey dots. All but two group means are significantly differentnt (Bonferroni corrected, Wilcoxon test, p < 0.05). Left column: entire 24-hour cycle (All); Central: daytime only; Right: nighttime only. Gold = diurnally variable baseline, blue = cold, orange = hot.ot. The only two insignificant comparisons are labeled with a bracket.

1 bioRxiv preprint doi: https://doi.org/10.1101/2020.12.18.423523; this version posted December 21, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

Figure 3. 24-hour time series of energy expenditure (EE, left), respiratory quotient (RQ, center),r), and water loss (H2O, right) for males and females across the three experimental groups: yellow = baseline or normal diurnally variable environmental conditions, orange = constant hot, blue = constant cold. Black dashed lines depict scaled daily temperature variation across experiments and solid lines (RQ plots only) indicate the food quotient (0.89).

2 bioRxiv preprint doi: https://doi.org/10.1101/2020.12.18.423523; this version posted December 21, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

Figure 4. Change points (hr:min), denoting shifts in mean and variance for EE (left), RQ (center), and H2O (right) across a 24 hour cycle for females (top 3 rows) and males (bottom 3 rows) for each experiment (baseline = yellow, hot = orange, cold = blue). Vertical, dashed bars indicate change point estimates, with light grey indicating over fit change points. Light grey shading indicates nighttime (e.g., dark photoperiod), dark grey shading indicate transition periods (only present in baseline experimental treatments), and no colored shading indicates daytime (e.g., light photoperiod).

3 bioRxiv preprint doi: https://doi.org/10.1101/2020.12.18.423523; this version posted December 21, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

TABLES Table 1. Summary statistics (median = med., mean, standard deviation = s.d., range) for females (top) and males (bottom) for each response variable: energy expenditure (EE), water

loss (H2O, mg), respiratory quotient (RQ), carbon-dioxide produced (VCO2), and oxygen

consumed (VO2) across each experiment (Baseline, Cold, Hot) during the daytime (rest) and nighttime (active) periods. EE H2O (mg) RQ Data med. mean sd range med. mean sd range med. mean sd range All 0.15 0.17 0.10 0.02-0.53 0.14 0.15 0.08 -0.11-0.42 0.91 0.92 0.12 0.51-1.50 Baseline 0.12 0.15 0.10 0.02-0.46 0.12 0.12 0.07 -0.11-0.37 0.94 0.97 0.18 0.51-1.50 Day 0.08 0.09 0.05 0.03-0.32 0.13 0.14 0.06 0.03-0.37 1.05 1.05 0.17 0.62-1.49 Night 0.23 0.24 0.09 0.04-0.46 0.09 0.08 0.06 -0.11-0.24 0.87 0.86 0.12 0.52-1.36 Cold 0.21 0.23 0.09 0.06-0.53 0.12 0.13 0.06 -0.03-0.35 0.92 0.91 0.07 0.67-1.23 Day 0.18 0.19 0.06 0.06-0.48 0.09 0.10 0.05 -0.03-0.27 0.92 0.91 0.07 0.67-1.18

Females Females Night 0.27 0.28 0.10 0.06-0.53 0.15 0.16 0.06 0.04-0.34 0.93 0.92 0.05 0.74-1.23 Hot 0.11 0.13 0.07 0.02-0.42 0.18 0.19 0.08 -0.04-0.42 0.88 0.88 0.09 0.51-1.39 Day 0.09 0.10 0.05 0.02-0.39 0.16 0.16 0.08 -0.04-0.41 0.89 0.89 0.10 0.51-1.39 Night 0.17 0.18 0.08 0.02-0.41 0.22 0.23 0.08 0.04-0.42 0.87 0.88 0.07 0.69-1.29 All 0.17 0.19 0.11 0.03-0.60 0.15 0.16 0.08 -0.09-0.43 0.92 0.92 0.10 0.58-1.37 Baseline 0.14 0.19 0.13 0.03-0.54 0.16 0.16 0.10 -0.09-0.43 0.90 0.93 0.12 0.65-1.37 Day 0.09 0.10 0.06 0.03-0.49 0.18 0.19 0.08 0.05-0.43 1.00 1.01 0.12 0.70-1.37 Night 0.31 0.31 0.10 0.06-0.52 0.12 0.11 0.09 -0.09-0.39 0.85 0.85 0.06 0.65-1.22 Cold 0.24 0.25 0.10 0.10-0.60 0.14 0.15 0.07 -0.05-0.39 0.91 0.90 0.07 0.71-1.18 Day 0.19 0.21 0.06 0.10-0.57 0.11 0.11 0.04 -0.03-0.24 0.91 0.90 0.06 0.72-1.18 Males Males Night 0.31 0.32 0.10 0.12-0.60 0.18 0.19 0.07 -0.04-0.39 0.92 0.91 0.07 0.71-1.15 Hot 0.11 0.13 0.08 0.03-0.47 0.17 0.17 0.08 -0.04-0.43 0.94 0.94 0.10 0.58-1.34 Day 0.08 0.09 0.04 0.03-0.27 0.13 0.13 0.05 -0.04-0.29 0.94 0.95 0.10 0.66-1.34 Night 0.18 0.19 0.10 0.04-0.47 0.23 0.23 0.07 0.04-0.43 0.93 0.93 0.09 0.58-1.30 VCO2 VO2 Data med. mean sd range med. mean sd range All 0.45 0.52 0.29 0.06-1.59 0.51 0.58 0.33 0.05-1.82 Baseline 0.38 0.46 0.28 .07-1.59 0.41 0.51 0.34 0.05-1.59 Day 0.26 0.29 0.15 0.09-1.12 0.24 0.28 0.16 0.08-1.05 Night 0.64 0.69 0.26 0.10-1.39 0.78 0.81 0.30 0.14-1.59 Cold 0.65 0.69 0.28 0.17-1.59 0.71 0.76 0.30 0.19-1.82 Day 0.55 0.58 0.20 0.17-1.51 0.62 0.64 0.21 0.19-1.59

Females Females Night 0.82 0.87 0.30 0.17-1.59 0.89 0.95 0.33 0.19-1.82 Hot 0.33 0.39 0.22 0.06-1.24 0.38 0.45 0.25 0.06-1.44 Day 0.26 0.30 0.14 0.08-1.13 0.31 0.33 0.16 0.08-1.32 Night 0.49 0.53 0.24 0.06-1.24 0.57 0.61 0.28 0.06-1.38 All 0.51 0.59 0.34 0.09-1.82 0.57 0.65 0.39 0.09-2.03 Baseline 0.43 0.58 0.36 0.10-1.62 0.48 0.66 0.43 0.09-1.85 Day 0.29 0.33 0.17 0.10-1.54 0.29 0.34 0.19 0.09-1.62 Night 0.90 0.89 0.30 0.10-1.57 1.07 1.04 0.33 0.20-1.79 Cold 0.72 0.76 0.29 0.34-1.82 0.80 0.85 0.32 0.33-2.03 Day 0.59 0.64 0.18 0.34-1.71 0.66 0.71 0.21 0.33-1.95 Males Males Night 0.95 0.97 0.30 0.36-1.82 1.05 1.08 0.34 0.39-2.03 Hot 0.34 0.41 0.26 0.09-1.42 0.36 0.45 0.28 0.10-1.58 Day 0.27 0.29 0.12 0.10-0.80 0.28 0.31 0.13 0.10-0.90 Night 0.56 0.60 0.30 0.13-1.42 0.61 0.65 0.32 0.12-1.58

4 (which wasnotcertifiedbypeerreview)istheauthor/funder,whohasgrantedbioRxivalicensetodisplaypreprintinperpetuity.Itmade bioRxiv preprint

Table 2. Maximum-likelihood change point estimates for males and females, across each experiment (baseline, hot, cold), based on AIC optimization. True change points in black, overfit change points in grey italics. Cold environmental conditions are denoted in blue, hot in orange, and environmental transitions (baseline only) are shown in yellow. doi: https://doi.org/10.1101/2020.12.18.423523

Time (hr:min) 22 16 23 Sex Exp. DV 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 17 18 19 20 21 24 EE 5:08 6:52 7:36 18:34 19:18 20:20 RQ 6:42 19:52

H2O 6:50 11:48 20:00 available undera

Baseline Baseline VCO2 5:08 6:52 7:36 18:34 19:18 20:12 VO2 5:34 6:52 7:36 18:34 19:18 20:20 EE 7:34 13:53 20:55 RQ 8:53 15:40 H2O

7:36 21:31 CC-BY-NC-ND 4.0Internationallicense Hot Hot

Females VCO2 7:34 13:53 20:55 ; VO2 7:34 13:53 20:55 this versionpostedDecember21,2020. EE 6:20 19:56 RQ 6:14 15:12 H2O 6:44 14:52 20:28 Cold Cold VCO2 6:14 19:56 VO2 6:20 19:56 EE 5:39 6:53 8:12 18:56 20:12 RQ 5:49 7:58 12:16 19:26 20:30 H2O 3:20 7:21 11:46 14:46 20:23

Baseline Baseline VCO2 5:39 6:53 8:13 18:59 20:12

VO2 5:39 6:53 8:12 18:56 20:12 . EE 7:28 10:32 20:46 s RQ 12:06 21:05 The copyrightholderforthispreprint le a H2O 7:28 10:10 21:23 M Hot VCO2 1:57 6:53 8:59 20:46 VO2 7:28 10:32 20:46 EE 5:56 20:06 RQ 11:04 H2O 7:35 20:52 Cold Cold VCO2 5:56 20:06 VO2 5:56 20:06

5 (which wasnotcertifiedbypeerreview)istheauthor/funder,whohasgrantedbioRxivalicensetodisplaypreprintinperpetuity.Itmade bioRxiv preprint

doi: https://doi.org/10.1101/2020.12.18.423523 available undera CC-BY-NC-ND 4.0Internationallicense ; this versionpostedDecember21,2020. . The copyrightholderforthispreprint

6